cellxgene-census

Query single-cell expression data across tissues and diseases using programmatic access to the CELLxGENE Census. You can filter by cell metadata, tissue types, and disease states to retrieve scalable, reproducible subsets of data for cross-dataset analysis.

Yes, AnnData and Scanpy integrate with CELLxGENE Census for downstream analysis. You can run an in-memory query for specific cell types and tissues, returning an AnnData object directly compatible with Scanpy workflows for visualization and processing.

Use out-of-core querying with axis_query to process millions of single-cell observations in chunks. This scalable workflow prevents memory exhaustion by enabling chunked processing of expression data directly from the CELLxGENE Census.

Yes, specifying a census_version is required for reproducibility when querying CELLxGENE Census. It ensures that your programmatic access to single-cell expression data returns consistent results across different runs and analysis environments.

Apply the is_primary_data filter along with tissue_general and feature_id filters when querying single-cell datasets. These metadata filters ensure you retrieve only primary, non-duplicated observations for your specific tissue and gene targets.

Yes, you can train PyTorch models using multi-dataset single-cell expression data. The CELLxGENE Census supports multi-dataset analysis, allowing you to query and integrate cross-tissue comparisons and expression data for model training.